BTW, nowadays everyone talks about AI, machine learning and "mobile-first". But when I open any mobile app or any mobile website with ads I see only ads of "clash of kings" and similar scammy games. They collect lots of data but ads have no targeting at all. At least ads on mobile phones. I can't understand it.
I used to work in the industry - mobile ads can be quite targeted. My guess is apps like clash of clans appeal to a wide range of people and are backed by heavy ad spending. This means a wide variety of people will be targeted with their ads.
Doesn't mean ads arent targeted. E.g. a 25 year old white programmer probably isn't getting Spanish language ads, or ads targeting new mothers, or ads for retirement communities.
> Doesn't mean ads arent targeted. E.g. a 25 year old white programmer probably isn't getting Spanish language ads, or ads targeting new mothers, or ads for retirement communities.
You'd be surprised. I bought a travel sewing kit five years ago on Amazon, and ever since I'm getting advertisements and "recommendations" for handbags, makeup and high-heeled shoes. It's so blatantly sexist and wrong it's almost funny again.
> It's so blatantly sexist and wrong it's almost funny again.
Have you considered that it's blatantly sexist and right? I.e., that perhaps no-one programmed the ad network to associate sewing kits with handbags, makeup & high-heeled shoes, that perhaps the ad serving AIs learnt that on their own?
I wonder what we'll do when our AIs come to socially-unacceptable-but-true conclusions. Humans can be brow-beaten or persuaded into ignoring the truth systematically, but computers have to either have each bit of truth-denying programmed into them, or have much better intelligence and spend much more CPU calculating at a higher level in order to avoid socially-unacceptable truths.
> Have you considered that it's blatantly sexist and right?
I buy an average of a hundred items on Amazon a year, among them all my – male – clothes. Your algorithms are just plain shit when a single purchase five years ago is somehow weighed more than the whole rest.
So the algorithm isn't even able to differentiate between the buying behaviour of a married couple and a single male living alone, with roughly 10 years worth of buying history to judge from?
I'd fire the department responsible for that waste of money.
I buy my wife and daughter gifts using Amazon. I don't think its unreasonable or sexist or evil for an advertiser to assume I'll continue to buy gifts for my wife and daughter.
If the algorithm makes money I don't see the problem.
I mark everything I buy on Amazon as a gift. This seems to stop the silly recommendation behaviour, and also stops Amazon emailing you asking for reviews or to provide answers to other customers' questions.
Hmm, isn't "being sexist" the whole raison d'etre for recommendation algorithms? I mean, sexism or other form of chauvinism are basically estimation of individual traits based on group affiliation.
Is it though? Just because you aren't interested in these other recommendations and connected ads doesn't mean the wide majority who purchases sewing kits aren't as well.
If this "blantantly sexist" ads were wrong no one would purchase such ads. Sure they may be stereotypical, but that's what targeted ads are all about, creating data driven stereotypes that later can be used for increasing sales and relativity.
But they are wrong. The GP is not interested on those.
Yes, the "blatantly sexist" bias may have happened due to unbiased empiricism. But the ads as still wrong. Id the GP was an exception, that would be an worthless anecdote, but anecdotes of ads being correct are hard to find.
You can fix that. If you go to Amazon -> Your Account -> Your Recommendations -> Improve My Recommendations, you can tell it to ignore certain purchases.
I have to do that surprisingly frequently, myself. I buy things buy as gifts, as one off experiments, or with no intention of using them for their intended purposes. ;)
Most ads-supported games have only ads of ad-supported games. That would be sane if it was a fast growing market with reasonable barriers to entry, but "free" games are not.
